Horizon Robotics unveils new AI embedded smart camera

Dr. Kai Yu of Horizon Robotics

AI startup Horizon Robotics introduced a new HD smart camera with embedded AI that is capable of facial recognition at up to 99.7% accuracy. The smart camera HR-IPC2143 is a first of its kind with applications in public security for the real-time detection of potential suspects, and in retail for identifying VIP customers as they enter a store. The device was announced at the International Security Technology Show (ISC West 2018), the largest physical security exhibition in North America.

“The smart camera is a reflection of Horizon Robotics’ mission—to provide high-performance, low-power and low-cost products utilising cutting-edge artificial intelligence technology,” said Dr. Kai Yu, CEO and founder of Horizon Robotics. “Other companies can only deliver a portion of the equation, but we are able to integrate all these form factors to create a complete business solution rather than just a stand-alone camera.”

The new Horizon Robotics smart camera performs local-capture and recognition capabilities simultaneously, providing facial capture, feature extraction and facial feature comparison based on deep-learning algorithms. The camera relies on a unique “algorithm + chip + cloud” system architecture that empowers users to process and convert the unstructured video data in the camera to achieve multi-target location detection, multi-target tracking, multi-target recognition and pedestrian attribute classification.

The camera supports high-performance facial recognition with a 50,000 feature library and processes 1920×1080p@30fps videos in real time. It can detect, track and identify 200 objects simultaneously within each frame.

For public safety, the camera offers optimal surveillance control, capable of recognising faces in crowded pathways, entry and exit ways, and at large-scale events. At the front end, it can read each video frame and conduct high-performance and power-saving facial capture and recognition, and perform related attribute analysis and structured interpretation of videos.

The same technology is applicable in a retail or commerce setting, as the system can recognise shoppers when they enter a brick-and-mortar store. A frequent shopper or big spender can be quickly identified so store staff can provide special treatment, including VIP offers or discounts.

The camera employs the “Sunrise” embedded AI computer vision processor, which features the Brain Processing Unit (BPU) architecture. BPU is a proprietary and high-performance AI processor architecture independently developed by Horizon to combine AI algorithms and hardware design. With a processing power of up to 1 Tops, the chip ensures that the small-body camera achieves the capture and recognition functions at the local end while improving the timeliness and stability of the camera. It also effectively reduces bandwidth cost.

Compared with previous smart cameras, the ability to capture and identify images simultaneously is a major breakthrough in the field of embedded AI in security and commercial applications, raising the development of the embedded AI industry to a whole new level.

Comment on this article below or via Twitter: @IoTNow OR @jcIoTnow

Recent Articles

realme CEO unveils 2021 plans following international awards

Posted on: January 15, 2021

According to the CEO of realme, though challenging year 2020 saw “positives on many fronts for realme, from corporate to product to design and more”. realme cemented its position as a mainstream global smartphone brand, ranking 7th overall in terms of shipments.

Read more

Where old meets IoT, SaaS integration

Posted on: January 15, 2021

Old meets new in the Internet of Things (IoT). It has been seen that we as humans only look forward when supported by evidence we already know (the past). This means, says Joseph Zulick of MRO Electric and Supply, that it becomes harder to find solutions that we have not already considered.

Read more